polar orbiting satellite data from 1979 onwards using CO2 slicing to infer cloud amount and height. More recently this has been continued with the MODIS data ...
USING CO2 SLICING TO INFER GLOBAL CLOUD COVER (IMPROVING ALGORITHM APPLICATIONS AND OVERCOMING INSTRUMENT IDIOSYNCRASIES) W. Paul Menzela , Bryan Bauma, Darren L. Jacksonb, Richard Freya, Elisabeth Weisza, Don Wyliea, Erik Olsona, and John Batesc a
b
Space Science and Engineering Center 1225 West Dayton Street University of Wisconsin -Madison, Madison, WI, USA Madison, WI, USA Cooperative Institute for Research in Environmental Sciences Boulder, CO c National Climatic Data Center, NOAA NESDIS Asheville, NC
Abstract The frequency of occurrence of upper tropospheric clouds has been extracted from NOAA/HIRS polar orbiting satellite data from 1979 onwards using CO2 slicing to infer cloud amount and height. More recently this has been continued with the MODIS data on the Terra and Aqua platforms. Algorithm adjustments for instrument noise, sensor to sensor differences, viewing angle, spectral response shifts, calculated versus measured radiance biases, changing CO2 and O3 amounts, and investigator error have been studied. CALIPSO measurements are being used to verify the improvements in the CO2 slicing algorithm. Some lessons learned are being documented and the HIRS/MODIS trend analyses are being compared with ISCCP. This presentation will touch on all of these topics.
HIRS CLOUD OBSERVATIONS SINCE EARLY 1980S Clouds are a strong modulator of amount of solar heating and thermal cooling of the earth system. We are currently in a warming trend that is often attributed to the increase in carbon dioxide (CO2) in the atmosphere. But clouds cover nearly three-fourths of the earth and a small increase in cloud cover could offset the warming from increased CO2 in the atmosphere. It has been found that clouds vary greatly over the earth, just as our weather does, and their effect on warming and cooling has to be inferred from the combined effect of all clouds in all places. Darker thicker clouds are cooling the earth through their reflection of sunlight. Thin ice clouds, called cirrus, allow sunlight to enter the earth system but trap thermal radiation attempting to leave. Cirrus clouds are warming the earth. The International Satellite Cloud Climatology Program (ISCCP) has collected the largest global cloud data set using visible and infrared measurements from the international suite of weather satellites. As a supplement multi-spectral infrared measurements from the National Oceanic and Atmospheric Administration polar orbiting High Resolution Infrared Radiometer Sounders (HIRS) have been used for enhanced cirrus detection. Using regions of the infrared spectrum with differing sensitivity to atmospheric carbon dioxide, the HIRS measurements probe the atmosphere to different depths and reveal thin ice clouds high in the atmosphere. Since 1979, HIRS measurements have found clouds most frequently in two locations; (1) the Inter-Tropical Convergence Zone (ITCZ) in the deep tropics where trade winds converge and (2)
the middle to high latitude storm belts where low pressure systems and their fronts occur. In between are latitudes with fewer clouds and rain called sub-tropical deserts over land and subtropical high pressure systems over oceans. The decadal average cloud cover has not changed appreciably from the 1980s to the 1990s. Small increases occurred in the tropics, mainly in the Indonesian Islands. Small decreases occurred in the sub-tropics, the eastern Sahara and in the central Pacific Ocean from Hawaii westward. The decreasing trend in Antarctica is uncertain because cloud detection itself is very difficult in the cold temperatures of Antarctica. High cloud cover has changed some in the northern hemisphere winter season. Increases of 10% in the last decade for clouds above 6 km altitude occurred in the western Pacific, Indonesia, and over Northern Australia. Other fairly large increases occurred in western North America, Europe, the Caribbean, Western South America, and the Southern Ocean north of Antarctica. Decreases in high clouds occurred mainly in the tropical South Pacific, Atlantic and Indian Oceans south of the ITCZ. Figure 1 shows the globally averaged frequency of cloud detection (excluding the poles where cloud detection is less certain) has stayed relatively constant at 75%; there are seasonal fluctuations but no general trends. High clouds in the upper troposphere (above 6 km) are found in roughly one third of the HIRS measurements; a small increasing trend of ~ 2% per decade is evident.
Figure 1: The monthly average frequency of clouds and high clouds (above 6 km) from 70 south to 70 north latitude from 1979 to 2002. The most significant feature of these data may be that the globally averaged cloud cover has shown little change in spite of dramatic volcanic and El Nino events. During the four El Nino events winter clouds moved from the western Pacific to the Central Pacific Ocean, but their global average in the tropics did not change. El Chichon and Pinatubo spewed volcanic ash into the stratosphere that took 1-2 years to fall out, but cloud cover was not affected significantly. These cloud data show overall constancy, with a small increase in high clouds. The high thin clouds capture some of the earth’s infrared radiation similar to CO2 and thus they contribute to global warming in the same manner. Clouds do not appear to be off-setting global warming by increasing their reflection of incoming solar radiation; they are possibly enhancing it with modest increase of thermally trapping high thin ice clouds.
The time series reported here were taken from the NOAA satellites gaps in the 8 am/pm orbit precluded their inclusion. Measurements and the increase in atmospheric CO2 (see Wylie et al., 2005). removed from the data record; subsequent reprocessing of the HIRS for sensor to sensor differences in spectral response functions.
in the 2 am/pm orbit; data were adjusted for orbit drift Anomalous sensors were data will attempt to account
CO2 SLICING ALGORITHM TESTS USING MODIS COMPARISONS WITH CALIPSO Cloud properties derived from measurements with the CO2 sensitive bands and the infrared window on MODIS have been compared with co-located CALIPSO (Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation) observations. The CALIPSO instrument takes measurements about 75 seconds behind that of Aqua MODIS as they both orbit the Earth from pole to pole. Collocation is accomplished by matching the CALIPSO latitude and longitude to those of a 5x5 km MODIS cloud top pressure. MODIS cloud top pressures were converted to heights using the Global Forecast System and were compared to heights from analysis of CALIPSO 0.532 micron backscatter data. Various adjustments to the MODIS cloud algorithm used in Collect 5 (described in Menzel et al., 2008) were tested and compared with Calipso cloud observations. Figure 2a shows the histogram of the cloud height differences of over 2 million MODIS Collect 5 cloud products minus the Calipso cloud measurements distributed globally between 60 N and 60 S. In the mean Calipso is higher than MODIS by 2.7 km with a scatter of 4.1 km. High cloud determinations from MODIS are hindered by thin cloud heights defaulting to infrared window estimates; MODIS also errs on low marine stratus clouds by placing them above the inversion height. Several adjustments to the collect 5 algorithm were attempted. They include: Test 1: Allow the tropopause level to be the cloud height solution when no intersection is found between measured and the calculated terms in the CO2-slicing equation; lower and upper bounds for a solution are allowed to be the window channel solution and tropopause, respectively. Test 2: Perform selection of the final cloud top pressure by the "top-down” method; 36/35, 35/34, 34/33 in that order. Test 3: Lower the "noise" limits (clear vs. cloudy radiances required to be greater than a limit set for each of bands 33-36 to imply cloud presence). This impacts the number of observations processed by the CO2-slicing algorithm as opposed to simple IR window channel technique. The intent is to force CO2 slicing solutions more often for high thin clouds. Test 4: Adjust the input data so that the ozone profile between 10 and 100 hPa agrees with the values in the GDAS data set instead of using climatology. This is necessary because forward calculation of CO2 radiances is influenced by O3 profiles. Test 5: Prohibit CO2 slicing solutions for water clouds; use only IRW solution. Avoid IRW solutions for ice clouds; use CO2 slicing whenever possible. Restrict CO2 channel pair solutions to the appropriate portion of troposphere (determined by their weighting functions – 36/35 less than 400 hPa, 35/34 less than 550 hPa, and 34/33 less than 700 hPa). Test 5c: Implement the Band 34, 35, 36 spectral shifts suggested by Tobin et al. (2005). The results of all the tests are shown in Figure 2b. The mean cloud height difference dropped to 2.5 km with a scatter of 3.8 km. Cloud above 5 km showed the biggest improvement (mean difference dropping from 4.5 to 4.1 km with scatter reduced from 4.6 to 4.4 km). While these improvements are modest on the global scale, we believe that they represent significant improvements to the implementation of the CO2 slicing algorithm for MODIS and HIRS data.
CONCLUSIONS The algorithm tests with MODIS data are suggesting new strategies for reprocessing the HIRS data. Conclusions are:
* The largest cloud height differences results from not using CO2 slicing (>15 km). * Reducing the cloud detection threshold produced more high thin cloud retrievals, but also produced erroneously high CO2 CTH retrievals for low water clouds in the southern Pacific. * CO2 slicing (IRW) heights should be avoided for water (ice) clouds. * A high bias in marine stratus was identified in the MODIS retrievals; CTH algorithm problems in inversions will be mitigated assuming a wet lapse rate. * Adjusting the spectral response of the CO2 bands reduced CTH errors. * Selecting the cloud top pressure from spectral radiance ratio using a top down criteria improved high cloud detection. * Making multiple passes through large data sets was necessary. * Using CALIOP as a reference was invaluable.
REFERENCES Tobin, D. C., H. E. Revercomb, C. C. Moeller, and T. S. Pagano, 2006: Use of AIRS high spectral resolution infrared spectra to assess the calibration of MODIS on EOS Aqua, J. Geophys. Res., 111, D09S05, doi:10.1029/2005JD006095. Menzel, W. P., R. A. Frey, H. Zhang, D. P. Wylie., C. C. Moeller, R. A. Holz, B. Maddux, B. A. Baum, K. I. Strabala, and L. E. Gumley, 2008: MODIS global cloud-top pressure and amount estimation: algorithm description and results. Jour of App Meteor and Clim., 47, 1175-1198. Wylie, D. P., D. L. Jackson, W. P. Menzel, and J. J. Bates, 2005: Global Cloud Cover Trends Inferred from Two decades of HIRS Observations. Journal of Climate, Vol. 18, No. 15, pages 3021–3031.
Figure 2a: Global 60 N to 60 S normalized histogram of MODIS minus CALIPSO cloud top height measurements for all (black), higher than 5 km (red), and lower than 5 km (blue)
new
Figure 2b: New global 60 N to 60 S normalized histogram of MODIS minus CALIPSO cloud top height measurements for all (black), higher than 5 km (red), and lower than 5 km (blue) after implementing all of the tests described above.
